Hierarchical Bayesian model for diffuse optical tomography of human brains

Abstract

Diffuse optical tomography (DOT) is emerging technology to improve spatial resolution of conventional multichannel near infrared spectroscopy (NIRS). Although the scalp blood flow heavily contaminates the cerebral blood flow, all of previously proposed DOT algorithms fail to provide a way to segregate these two components. Here we propose a hierarchical Bayesian model and DOT reconstruction algorithm to segregate the cerebral blood flow from the scalp blood flow. The key idea of our method is that the different prior distributions for the scalp and cerebral blood flow are assumed based on observations that spatial distribution of scalp blood flow is broad whereas that of the cerebral blood flow is focal. Our DOT results were compared with fMRI data using human experimental data.